Scholarlink

<p>Artificial intelligence driven educational technologies have rapidly transformed multiple domains, particularly primary education and language learning, where interactive systems such as augmented reality platforms and conversational agents enhance personalized instruction and learner engagement. These technologies enable adaptive feedback, real-time interaction tracking, and multimodal content delivery, creating new opportunities for improving English vocabulary acquisition. However, despite rapid technological achievements, structured integration of multi-source learning datasets remains limited, restricting comprehensive analysis of cognitive and behavioural learning outcomes. To solve this problem, this paper proposes a comprehensive dataset collation and harmonization framework that integrates three publicly available datasets: AR-based English vocabulary learning, student learning interaction logs, and chatbot-based English learning data. The methodology includes four stages. First, heterogeneous datasets are standardized through unified student identifiers, session mapping, and timestamp normalization. Secondly, vocabulary accuracy and response variables are cleaned and aligned across sources. Thirdly, engagement-related metrics such as time-on-task, dialogue turns, and interaction frequency are computed to capture behavioural dimensions. Finally, derived indicators, including accuracy rate, average response latency, and engagement index, are generated to enable multivariate statistical modelling. Experimental results indicate that the integrated dataset improves feature completeness by over 35% compared to single-source datasets and enables stronger correlation modelling between engagement and vocabulary accuracy (<i>r</i> &gt; 0.45, <i>p</i> &lt; 0.01). Principal component analysis demonstrates clearer separation of performance clusters, explaining more than 70% of total variance within the first two components. Compared with isolated datasets, the proposed framework provides greater analytical robustness, richer feature representation, and enhanced predictive capability for modelling technology-enhanced vocabulary learning outcomes.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Research on the application of interactive social robot in english vocabulary learning in primary school

  • 期刊论文

  • 发表日期：

    2026年7月8日

查看全文
  • Suqin Wu,
  • Mingyong Pang,
  • Xuemei Sun,
  • Xinjian Wang

摘要

Artificial intelligence driven educational technologies have rapidly transformed multiple domains, particularly primary education and language learning, where interactive systems such as augmented reality platforms and conversational agents enhance personalized instruction and learner engagement. These technologies enable adaptive feedback, real-time interaction tracking, and multimodal content delivery, creating new opportunities for improving English vocabulary acquisition. However, despite rapid technological achievements, structured integration of multi-source learning datasets remains limited, restricting comprehensive analysis of cognitive and behavioural learning outcomes. To solve this problem, this paper proposes a comprehensive dataset collation and harmonization framework that integrates three publicly available datasets: AR-based English vocabulary learning, student learning interaction logs, and chatbot-based English learning data. The methodology includes four stages. First, heterogeneous datasets are standardized through unified student identifiers, session mapping, and timestamp normalization. Secondly, vocabulary accuracy and response variables are cleaned and aligned across sources. Thirdly, engagement-related metrics such as time-on-task, dialogue turns, and interaction frequency are computed to capture behavioural dimensions. Finally, derived indicators, including accuracy rate, average response latency, and engagement index, are generated to enable multivariate statistical modelling. Experimental results indicate that the integrated dataset improves feature completeness by over 35% compared to single-source datasets and enables stronger correlation modelling between engagement and vocabulary accuracy (r > 0.45, p < 0.01). Principal component analysis demonstrates clearer separation of performance clusters, explaining more than 70% of total variance within the first two components. Compared with isolated datasets, the proposed framework provides greater analytical robustness, richer feature representation, and enhanced predictive capability for modelling technology-enhanced vocabulary learning outcomes.